T-distributed Stochastic Neighbor Embedding Research Articles

Exploring the Educational Landscape of ChatGPT: A Topic Modeling Approach on Twitter Data

In the rapidly evolving landscape of Artificial Intelligence (AI), platforms like ChatGPT are reshaping the educational domain, prompting deeper explorations into the nature and depth of this intersection. This study aimed to systematically uncover the prevailing sentiments, concerns and discussions on Twitter surrounding ChatGPT’s role in education. Through an extensive data collection process, over 3.8 million tweets were initially gathered, followed by rigorous refining processes that included expert-driven tweet labelling and subsequent classification using Machine Learning (ML) and deep learning models. The cleaned dataset underwent a series of preprocessing steps and feature extraction and was ultimately subjected to Latent Dirichlet Allocation (LDA) for topic modelling. Our findings unveiled 15 distinct topics that broadly spanned common discussions, AI implementation, and its potential impacts. The data’s visualisation using t-distributed stochastic neighbour embedding (t-SNE) showcased a dense central clustering of these topics. In conclusion, our research underscores the multi-faceted dialogues on AI, particularly ChatGPT, in education, emphasising the pressing need for continued discourse and research as AI tools further integrate into our educational paradigms.

Real-time identification of precursors in commercial aviation using multiple-instance learning

This research pioneers the application of precursor concepts to preemptively identify and prevent aviation safety incidents using Machine Learning (ML). Airlines and governing organizations, such as the Federal Aviation Administration (FAA) in the United States, have been trying to prevent safety incidents during routine operations. However, this task is challenging due to the lack of timestep-wise event annotation in flights and the complexity involved in the timely identification of incidents prior to their occurrence. To address these issues, we propose a real-time precursor identification methodology combining Multiple-Instance Learning (MIL) and feature-based Knowledge Distillation (KD) learning. Our two-stage approach, involving deep MIL for labeling and a KD-based model for real-time warnings, demonstrates state-of-the-art performance and a time delay of 2.99ms using a dataset of 23,549 real flights. Further experiments using t-distributed Stochastic Neighbor Embedding (t-SNE) and occlusion method confirm our model’s transparency, enabling the generation of reliable quantitative precursor scores and facilitating reasoning about the causes of safety incidents at the parameter level. Additionally, statistical analysis of precursors reveals varying evolution times for different safety events, which indicates that pilots have at least 8 s to react after receiving a warning. In conclusion, our research provides a theoretical foundation and technical support for the next generation of online risk warning systems, enhancing flight safety and offering a pathway towards more intelligent and secure flight operations.

Identification method for micro–nano defects in Si3N4 ceramic bearing rollers

In response to the surface micro–nano defects of small Si3N4 ceramic bearing rollers during use, as well as the poor generalization ability and low accuracy of the object detection model trained on the micro and nano defect dataset, an enhanced recognition algorithm based on deep convolutional generative adversarial networks is proposed. Due to the limited size of the dataset, the DCGAN model is reconstructed to effectively expand the micro–nano defect dataset. In addition, noise generalization is applied to stabilize DCGAN model training, creating a low-dimensional manifold distribution to ensure significant overlap between the data and the original dataset, and activating Jensen–Shannon (JS) divergence for stable training. To verify the effectiveness of the enhanced dataset, synthetic micro–nano defects are used to improve the YOLO-v4-tiny object detection model. By comparing t-distributed stochastic neighbor embedding (t-SNE) and feature vectors, it can be found that the images generated by the optimized DCGAN have higher grayscale feature diversity and better visual consistency. After generations of enhancements, the micro–nano defect detection speed has reached 226FPS, and the accuracy has reached 97.41%.

Addressing large scale patterns of no-flow events in rivers: An in-depth analysis with Achelous software

Establishment of hydrological criteria that could serve as guidelines for addressing intermittency is not an easy task. However, efforts in the last years are yielding promising advancements in this direction. Scientists have been working to unravel the complexities of intermittency dynamics. In this study, we aimed to investigate the characteristics of naturally intermittent water systems by exploring the occurrence of no-flow events. A hydrological model was employed to generate streamflow data. Our analysis encompassed a thorough examination of nineteen flow regime metrics, estimated across 2064 subbasins, with 190 of these meeting the criterion for intermittency. A custom Python library was deployed to automate the quantification of no-flow events, aligning with the concept of critical thresholds. Using the probabilistic t-distributed Stochastic Neighbor Embedding technique to capture complex patterns, three clusters were emerged. The first one was characterized by a low probability of no-flow events and a small number of no-flow events per year, the second cluster was abundant in no-flow events and demonstrated a tendency towards longer annual recession time scales. The third cluster stands out due to the significant variance in the duration of no-flow events. Concerning the time variability of the no-flow events, we concluded that they predominantly occurred during August. Both long- and short-term quantification of no-flow events should be under consideration so as to harmonize the naturally intermittent waterways with the water use requirements and the potential consequences of not meeting them. Future research should prioritize the investigation of hydroecology and ecohydrology in relation to streamflow dynamics and ecosystem interactions. By doing so, we can elevate our comprehension of how intermittent water systems function and their significance within the broader ecological context.

A stethoscope-guided interpretable deep learning framework for powder flow diagnosis in cold spray additive manufacturing

Cold spray (CS) particle deposition, also known as cold spray additive manufacturing, has garnered great attention as an advanced additive manufacturing (AM) and surface deposition technology, facilitating rapid and scalable production of functional parts and surfaces in a solid-state manner. In CS, consistent and precise feeding of functional feedstock powders is crucial for achieving effective particle deposition. However, vibratory-based powder feeders often face challenges associated with powder delivery and powder segregation. This underscores the critical need for a precise diagnostic framework to effectively control powder flow during cold spraying. To this end, the present study proposes a powder flow monitoring framework for the CS process using a stethoscope sound-guided interpretable deep learning (IDL) model. Internal sound data from the vibrated powder feeder is collected through a stethoscope sensor to train a two-stage model. In the first stage, a convolutional autoencoder (CAE) is trained to build an unsupervised learning-based anomaly detector, which identifies classification thresholds based on the receiver operation characteristic curve. In the second stage, a convolutional neural network (CNN) model is trained as the powder flow diagnostic tool by considering process anomalies, namely (i) no powder flow; (ii) feeder clogging; and (iii) no gas flow. The results reveal that the stethoscope sound-guided model achieves a classification accuracy of 95% on the test set, significantly outperforming benchmark utilizing typical external-sound recording microphones in diagnosing CS powder flow. Furthermore, the model is visualized and interpreted by employing t-distribution stochastic neighbor embedding and integrated gradients techniques to enhance reliability of CS powder flow diagnosis. This research highlights the effectiveness of the stethoscope sound-guided IDL model for in situ powder flow monitoring and process diagnosis in the domain of cold spray additive manufacturing, contributing to effective particle deposition.

A cluster approach to matching the competences of data specialists with skills in demand on the labour market

This paper addresses the challenge of aligning the competences of data specialists with skills in demand on the labour market in the rapidly evolving field of data science. Using an open dataset of 3,744 IT job postings, the study applies K-means clustering to identify key skill groupings for data specialist positions. The optimal number of clusters is determined using the elbow method, resulting in four distinct clusters: Data Analyst & Engineer, Data Platform Engineer, Data Science & Engineering Specialist, and Cloud Data Engineer. The research methodology employs unsupervised learning techniques, specifically K-means clustering, to analyze the distribution of skills across job postings. The clusters are visualized using t-distributed Stochastic Neighbor Embedding (t-SNE), providing insights into the relationships between different skill sets. The study reveals that job titles do not always unambiguously define the required skills, emphasizing the importance of focusing on specific skill sets rather than job titles alone. To bridge the gap between specific subject competences academic programs and industry requirements, the paper proposes a novel approach for comparing the proportion of skills in job clusters with the proportion of professional competencies in academic programs. This method is demonstrated using the Information Systems and Technologies Master's program at Kherson State University as a case study. The chi-square test is applied to confirm the statistical similarity between the skill structure of the Data Science & Engineering Specialist cluster and the competency structure of the academic program. The findings highlight the importance of continuous adaptation of profile of academic program to meet evolving industry needs. The proposed approach provides a data-driven framework for universities to align their programs with labor market demands, potentially improving graduate employability in the data science field. The study also underscores the need for personalized learning paths that can be tailored to individual career goals and skill gaps. Future research directions include the development of an artificial intelligence system to form individualized educational trajectories based on the skills required for specific job clusters. This could further enhance the alignment between education and industry needs, preparing students more effectively for the dynamic data science job market.

MDSC-FSPPA-LCFF network for diagnosis of rolling bearing with multipoint fault

Abstract Faults that occur in rolling bearings during operation are complex and variable. While extensive research has been conducted on compound faults involving multiple components, studies on multiple faults in single component are relatively scarce. However, the occurrence of multiple faults in single component is a common phenomenon. To address the issues of difficulty in feature extraction, numerous network parameters, and slow computational speed, a multi-scale dynamic snake convolution with fast spatial pyramid pooling attention (MDSC-FSPPA) and lightweight comprehensive feature fusion (LCFF) network is proposed for multipoint fault diagnosis of rolling bearings. Firstly, multi-scale shallow feature extraction (MSFE) module is applied to extract the features from the original signals. Then, dynamic snake convolution (DSC) with FSPPA module is used to refine these features deeply. Subsequently, LCFF module is employed to reduce network parameters while still fully extracting fault features. Additionally, fault identification is obtained through the softmax function. Finally, the t-distributed stochastic neighbor embedding (t-SNE) method is utilized to visually demonstrate the fault classification performance of the proposed method. The experimental evaluation conducted on bearing datasets indicates that the proposed network exhibits excellent performance of multipoint fault diagnosis in rolling bearings.

Anoikis-related genes in breast cancer patients: reliable biomarker of prognosis

BackgroundBreast cancer (BC) is the most common cancer in women, and its progression is closely related to the phenomenon of anoikis. Anoikis, the specific programmed death resulting from a lack of contact between cells and the extracellular matrix, has recently been recognized as playing a critical role in tumor initiation, maintenance, and treatment. The ability of cancer cells to resist anoikis leads to cancer progression and metastatic colonization. However, the impact of anoikis on the prognosis of BC patients remains unclear.MethodThis study utilized data from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases to collect transcriptome and clinical data of BC patients. Anoikis-related genes (ARGs) were classified into subtypes A and B through consensus clustering. Subsequently, survival prognosis analysis, immune cell infiltration analysis, and functional enrichment analysis were performed for both subtypes. Using the Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis, a set of 10 ARGs related to prognosis was identified. Immune cell infiltration and tumor microenvironment analyses were conducted on these 10 ARGs to develop a prognostic model. Furthermore, single-cell data analysis and real-time polymerase chain reaction (RT-PCR) analysis were employed to study the expression of the 10 identified prognostic ARGs in BC cells.ResultsOne hundred thirty-five ARGs were identified as differentially expressed genes in the TCGA and GEO databases, with 42 of them associated with the survival prognosis of BC patients. Analyses involving Principal Component Analysis (PCA), t-Distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP) revealed distinct expression patterns of ARGs between types A and B. Patients in type A exhibited worse survival prognosis and lower immune cell infiltration compared to type B. Subsequent analyses identified 10 key ARGs (YAP1, PIK3R1, BAK1, PHLDA2, EDA2R, LAMB3, CD24, SLC2A1, CDC25C, and SLC39A6) relevant to BC prognosis. Kaplan–Meier analysis indicated that high-risk patients based on these ARGs had a poorer BC prognosis. Additionally, Cox regression analysis established gender, age, T (tumor), N (nodes), and risk score as predictive factors in a nomogram model for BC. The model demonstrated diagnostic value for BC patients at 1, 3, and 5 years. Decision curve analysis (DCA) verified the risk score as a reliable predictor of BC patient survival rates. Moreover, RT-PCR results confirmed differential expressions of YAP1, PIK3R1, BAK1, PHLDA2, CD24, SLC2A1, and CDC25C in BC cells, with SLC39A6, EDA2R, and LAMB3 showing low expression levels.ConclusionARGs markers can be used as BC biomarkers for risk stratification and survival prediction in BC patients. Besides, ARGs can be used as stratification factors for individualized and precise treatment of BC patients.

The air transport research society world conference: A data science-based literature review on the years 2014–2024

The Air Transport Research Society (ATRS) World Conference is one of the major venues for air transport research. The conference covers a wide range of research talks, practice/industrial sessions, and research workshop activities. In this paper, we perform a data-driven analysis of the research abstracts that have been accepted and presented at the conference since 2014. We have grouped the abstracts from the ten annual conferences using t-distributed stochastic neighbor embedding to map high-dimensional keyword vectors into a two-dimensional plane for clustering, analysis, and visualization. The major focus of our study concerns three directions. First, we provide a formal description of the actual research presented at the ATRS World Conference series by using methods from natural language processing and machine learning, leading to a data-driven classification consisting of 35 major subject categories. Second, we analyze the origin of main authors/presenters and their background, including their institutions and countries of origin. Third, we perform a network-driven analysis of co-authorships across abstracts to identify the role and importance of key researchers in the community. Finally, we provide an analysis of popular research topics indicated by authors when submitting their abstracts, and a set of major recommendations for future work, based on the insights obtained from our study.

Ultra-lightweight convolution-transformer network for early fire smoke detection

BackgroundForests are invaluable resources, and fire is a natural process that is considered an integral part of the forest ecosystem. Although fire offers several ecological benefits, its frequent occurrence in different parts of the world has raised concerns in the recent past. Covering millions of hectares of forest land, these fire incidents have resulted in the loss of human lives, wild habitats, civil infrastructure, and severe damage to the environment. Around 90% of wildland fires have been caused by humans intentionally or unintentionally. Early detection of fire close to human settlements and wildlife centuries can help mitigate fire hazards. Numerous artificial intelligence-based solutions have been proposed in the past decade that prioritize the detection of fire smoke, as it can be caught through remote sensing and provide an early sign of wildland fire. However, most of these methods are either computationally intensive or suffer from a high false alarm rate. In this paper, a lightweight deep neural network model is proposed for fire smoke detection in images captured by satellites or other remote sensing sources.ResultsWith only 0.6 million parameters and 0.4 billion floating point operations per second, the hybrid network of convolutional and vision transformer blocks efficiently detects smoke in normal and foggy environmental conditions. It outperforms seven state-of-the-art methods on four datasets, including a self-collected dataset from the “Moderate Resolution Imaging Spectroradiometer” satellite imagery. The model achieves an accuracy of more than 99% on three datasets and 93.90% on the fourth dataset. The t-distributed stochastic neighbor embedding of extracted features by the proposed model demonstrates its superior feature learning capabilities. It is remarkable that even a tiny occurrence of smoke covering just 2% of the satellite image area is efficiently detected by the model.ConclusionsWith low memory and computational demands, the proposed model works exceedingly well, making it suitable for deployment in resource constrained devices for forest surveillance and early fire smoke detection.

Efficient navigation of cargo-towing microswimmer in non-uniform flow fields

The vision of deploying miniature vehicles within the human body for intricate tasks holds tremendous promise across engineering and medical domains. Herein, optimal navigation of a cargo-towing swimmer under an applied zig-zag flow is studied by employing direct numerical simulations coupled with a deep reinforcement learning algorithm. Tasks include navigation in flow and shear-gradient directions. We initially explore combinations of state inputs, finding that optimal navigation necessitates swimmers to perceive hydrodynamics and alignment, surpassing reliance solely on hydrodynamic signals while considering their memories. Next, we study combinations of action spaces, allowing dynamic changes in swimming and/or rotational velocities by tuning B1 and C1 parameters of the squirmer model, respectively. By keeping both parameters fixed, cargo-towing swimmers demonstrate superior performance in the flow direction compared to swimmers without load due to tumbling movements influenced by shear flow. In the shear-gradient direction, swimmers without load outperform cargo-towing swimmers, with performance decreasing as load length increases. Across the combination of allowing B1 and C1 to change, the policies from solely dynamic B1 actions demonstrate superior navigation. The policies are then used as a showcase against naive cargo-towing and inert colloidal chains. A t-distributed stochastic neighbor embedding analysis reveals the complex interplay between perceived hydrodynamic signals and swimmer position. In the flow direction, swimmers align effectively with regions of maximum velocity, while in the shear-gradient direction, periodic transitions from minimum to maximum state values occur. Comparing pullers, pushers, and neutral swimmers, cargo-towing swimmers show a reversal in swimming velocity trends, with pullers outpacing neutral and pusher swimmers, irrespective of load lengths. Published by the American Physical Society 2024

Recognition of Forward Head Posture Through 3D Human Pose Estimation With a Graph Convolutional Network: Development and Feasibility Study.

Prolonged improper posture can lead to forward head posture (FHP), causing headaches, impaired respiratory function, and fatigue. This is especially relevant in sedentary scenarios, where individuals often maintain static postures for extended periods-a significant part of daily life for many. The development of a system capable of detecting FHP is crucial, as it would not only alert users to correct their posture but also serve the broader goal of contributing to public health by preventing the progression of chronic injuries associated with this condition. However, despite significant advancements in estimating human poses from standard 2D images, most computational pose models do not include measurements of the craniovertebral angle, which involves the C7 vertebra, crucial for diagnosing FHP. Accurate diagnosis of FHP typically requires dedicated devices, such as clinical postural assessments or specialized imaging equipment, but their use is impractical for continuous, real-time monitoring in everyday settings. Therefore, developing an accessible, efficient method for regular posture assessment that can be easily integrated into daily activities, providing real-time feedback, and promoting corrective action, is necessary. The system sequentially estimates 2D and 3D human anatomical key points from a provided 2D image, using the Detectron2D and VideoPose3D algorithms, respectively. It then uses a graph convolutional network (GCN), explicitly crafted to analyze the spatial configuration and alignment of the upper body's anatomical key points in 3D space. This GCN aims to implicitly learn the intricate relationship between the estimated 3D key points and the correct posture, specifically to identify FHP. The test accuracy was 78.27% when inputs included all joints corresponding to the upper body key points. The GCN model demonstrated slightly superior balanced performance across classes with an F1-score (macro) of 77.54%, compared to the baseline feedforward neural network (FFNN) model's 75.88%. Specifically, the GCN model showed a more balanced precision and recall between the classes, suggesting its potential for better generalization in FHP detection across diverse postures. Meanwhile, the baseline FFNN model demonstrates a higher precision for FHP cases but at the cost of lower recall, indicating that while it is more accurate in confirming FHP when detected, it misses a significant number of actual FHP instances. This assertion is further substantiated by the examination of the latent feature space using t-distributed stochastic neighbor embedding, where the GCN model presented an isotropic distribution, unlike the FFNN model, which showed an anisotropic distribution. Based on 2D image input using 3D human pose estimation joint inputs, it was found that it is possible to learn FHP-related features using the proposed GCN-based network to develop a posture correction system. We conclude the paper by addressing the limitations of our current system and proposing potential avenues for future work in this area.

Deep learning based approach: automated gingival inflammation grading model using gingival removal strategy

Gingival inflammation grade serves as a well-established index in periodontitis. The aim of this study was to develop a deep learning network utilizing a novel feature extraction method for the automatic assessment of gingival inflammation. T-distributed Stochastic Neighbor Embedding (t-SNE) was utilized for dimensionality reduction. A convolutional neural network (CNN) model based on DenseNet was developed for the identification and evaluation of gingival inflammation. To enhance the performance of the deep learning (DL) model, a novel teeth removal algorithm was implemented. Additionally, a Grad-CAM + + encoder was applied to generate heatmaps for computer visual attention analysis. The mean Intersection over Union (MIoU) for the identification of gingivitis was 0.727 ± 0.117. The accuracy rates for the five inflammatory degrees were 77.09%, 77.25%, 74.38%, 73.68% and 79.22%. The Area Under the Receiver Operating Characteristic (AUROC) values were 0.83, 0.80, 0.81, 0.81 and 0.84, respectively. The attention ratio towards gingival tissue increased from 37.73% to 62.20%, and within 8 mm of the gingival margin, it rose from 21.11% to 38.23%. On the gingiva, the overall attention ratio increased from 51.82% to 78.21%. The proposed DL model with novel feature extraction method provides high accuracy and sensitivity for identifying and grading gingival inflammation.

Polymer design with enhanced crystallization tendency aided by machine learning

Designing the materials with desirable properties is very difficult task. Experimental approaches are expensive and time consuming. Machine learning (ML) guided screening is better option. In present study, different machine learning models are tried for the prediction of crystallization tendency of polymers. Hist gradient booting model is the best one. A large database of polymers is mined and easily synthesizable polymers are selected. Their crystallization tendency is predicted using fast ML model. A selected portion is analyzed using t-distributed stochastic neighbor embedding (t-SNE) method. Change in crystallization tendency on structural change is studied using structure Activity Landscape Index (SALI) analysis. On the selected polymers, clustering analysis is also performed to explore the structurally closely associated polymers with higher crystallization tendency. The synthetic accessibility assessment is also done for selected polymers. Our proposed approach is very useful for virtual screening of efficient materials.

Finding common development paths in voluntary national reviews reporting on sustainable development goals using aspect-based sentiment analysis.

Voluntary National Reviews (VNRs) provide a platform for participating countries to share their experiences, failures, and successes in achieving the United Nations (UN) Sustainable Development Goals (SDGs). The objective of this study is to gain a deeper understanding of the narrative elements, particularly the sentiment, in VNRs in order to more effectively assess and support global SDG progress. A total of 232 VNRs from 166 countries are analyzed using Aspect-Based Sentiment Analysis (ABSA) to extract each country's sentiment toward the 17 SDGs. The sentiment scores are then compared to the corresponding official UN SDG scores, and countries are grouped by their sentiment toward all 17 SDGs to identify potential common development pathways. The analysis uncovers a notable positive correlation between the reported sentiment and official SDG scores for SDG 2 (zero hunger) and SDG 11 (sustainable cities and communities), and a negative correlation for SDG 5 (gender equality). Conversely, this relationship is not significant for the majority of SDGs, suggesting that VNR narratives may not directly reflect actual progress. A t-distributed stochastic neighbor embedding (t-SNE) approach indicates a consistent sentiment score among developed countries. In contrast, there are greater differences in reporting sentiment among Emerging Markets, Frontier Markets, and Least Developed Countries (LDCs), where there is greater dispersion (especially among LDCs) and sentiment in reporting on SDG progress that appears to have changed from one reporting year to another. These findings highlight the need to interpret VNRs in the context of each country's unique situation and challenges specific to each country.

Label-free and liquid state SERS detection of multi-scaled bioanalytes via light-induced pinpoint colloidal assembly

Surface-enhanced Raman scattering (SERS) has been extensively applied to detect complex analytes due to its ability to enhance the fingerprint signals of molecules around nanostructured metallic surfaces. Thus, it is essential to design SERS-active nanostructures with abundant electromagnetic hotspots in a probed volume according to the dimensions of the analytes, as the analytes must be located in their hotspots for maximum signal enhancement. Herein, we demonstrate a simple method for detecting robust SERS signals from multi-scaled bioanalytes, regardless of their dimensions in the liquid state, through a photothermally driven co-assembly with colloidal plasmonic nanoparticles as signal enhancers. Under resonant light illumination, plasmonic nanoparticles and analytes in the solution quickly assemble at the focused surface area by convective movements induced by the photothermal heating of the plasmonic nanoparticles without any surface modification. Such collective assemblies of plasmonic nanoparticles and analytes were optimized by varying the optical density and surface charge of the nanoparticles, the viscosity of the solvent, and the light illumination time to maximize the SERS signals. Using these light-induced co-assemblies, the intrinsic SERS signals of small biomolecules can be detected down to nanomolar concentrations based on their fingerprint spectra. Furthermore, large-sized biomarkers, such as viruses and exosomes, were successfully detected without labels, and the complexity of the collected spectra was statistically analyzed using t-distributed stochastic neighbor embedding combined with support vector machine (t-SNE + SVM). The proposed method is expected to provide a robust and convenient method to sensitively detect biologically and environmentally relevant analytes at multiple scales in liquid samples.

Exploratory factor analysis for machining error data of compressor blades based on dimensionality-reduced statistical analysis method

Abstract Various machining errors inevitably occur on aero-engine compressor blades, including leading-edge contour error, trailing-edge contour error, camber contour error, and more. The current complexity surrounding the numerous ma-chining error types and their obscure interrelationships imposes immense effort for aerodynamic analysis and hin-ders overall error control. Thus, elucidating error correlations to achieve error dimensionality reduction is imperative. This study pioneers a dimensionality reduction approach via factor analysis to conduct a comprehensive statistical analysis of 13 types of blade machining errors. The proposed technique can categorize the 13 errors into three groups, each dominated by a distinct common factor. To validate the accuracy of the extracted factors, the factor scores of the three identified latent factors are computed. Cluster analysis is then performed on the factor scores, and the clustering results are visualized by t-Distributed Stochastic Neighbor Embedding (t-SNE). Furthermore, boot-strap resampling establishes the 95% confidence intervals for the factor scores. Capitalizing on the grouping struc-ture uncovered by factor analysis, multiple linear regression models are built for the errors within each group, and then, a preliminary conjecture is made about the potential key error types for each group of errors based on the re-gression coefficients. This hypothesis is then evidenced by the statistical analysis of cross-section profile error data of 28 blades. The present work can not only optimize machining processes but also relax tolerance requirements and diminish the effort of aerodynamic analysis.

Transformer fault diagnosis based on DBO-BiLSTM algorithm and LIF technology

Abstract In response to the deficiencies of traditional power transformer fault detection techniques, such as low sensitivity and the inability for online monitoring, a novel transformer fault diagnosis model combining Laser-Induced Fluorescence (LIF) technology with deep learning is proposed. Initially, the spectral data of transformer insulation oil is acquired using LIF technology, yielding spectral data for various fault types. Subsequently, MinMaxScaler (MMS) and Standard Normalized Variate (SNV) methods are employed for denoising and preprocessing the spectral data. The preprocessed data is then subjected to dimensionality reduction using Linear Discriminant Analysis (LDA) and T-distributed Stochastic Neighbor Embedding (T-SNE) to ensure that the spectral data retains maximal feature information while minimizing its dimensionality. Following this, Long Short-Term Memory (LSTM), Bi-directional Long Short-Term Memory (BiLSTM), Dung Beetle Optimizer-Bi-directional Long Short Term Memory (DBO-BiLSTM), Convolutional Neural Network (CNN), and Support Vector Machine (SVM) models are constructed. The reduced-dimensional data is fed into each of the five models for training to facilitate transformer fault diagnosis. Through comparative analysis among the five models, the optimal model is selected. Experimental results indicate that the DBO-BiLSTM model is the most suitable for transformer fault diagnosis in this experiment, underscoring its significant implications for ensuring the safety of power systems.

Combining Bulk and Single Cell RNA-Sequencing Data to Identify Hub Genes of Fibroblasts in Dilated Cardiomyopathy.

Dilated cardiomyopathy (DCM) is the second leading cause of heart failure, with intricate pathophysiological underpinnings. In order to shed fresh light on the mechanistic research of DCM, we combined bulk RNA-seq and single-cell RNA-seq (scRNA-seq) data to examine significant cells and genes implicated in the disease. This analysis employed publicly accessible bulk RNA-seq and scRNA-seq DCM datasets. The scRNA-seq data underwent normalization, principal component, and t-distribution stochastic neighbor embedding analysis. Cell-to-cell communication networks and activity analysis were conducted using CellChat. Utilizing enrichment analysis, the marker genes' role in the active cells was evaluated. After screening by limma software and weighted gene co-expression network analysis, the differentially expressed genes (DEGs) served as hub genes. Furthermore, these hub genes were subjected to immunological studies, transcription factor expression, and gene set enrichment. Lastly, the expression of the four hub genes and their connection to DCM were verified using the rat models. Fibroblasts and monocytes were chosen as hub cells from among the eight identified cell clusters; their marker genes intersected with DEGs to yield six hub genes. In addition, the six hub genes and the essential module genes intersected to yield four essential genes (ASPN,SFRP4,LUM, andFRZB) that were connected to the Wnt signaling pathway and highly expressed in fibroblast. The four hub DEGs had an expression pattern in the DCM rat model experiment results that was in line with the findings of the bioinformatics study. Additionally, there was a strong correlation between decreased cardiac function and the up-regulation of ASPN,SFRP4,LUM, and FRZB. Ultimately, bulk RNA-seq and scRNA-seq data identified fibroblasts and monocytes as the main cell types implicated in DCM. The highly expressed genes ASPN,FRZB,LUM,and SFRP4 in fibroblasts may aid in the mechanistic investigation of DCM.

A Complete Scheme for Multi-Character Classification Using EEG Signals From Speech Imagery.

Some classification studies of brain-computer interface (BCI) based on speech imagery show potential for improving communication skills in patients with amyotrophic lateral sclerosis (ALS). However, current research on speech imagery is limited in scope and primarily focuses on vowels or a few selected words. In this paper, we propose a complete research scheme for multi-character classification based on EEG signals derived from speech imagery. Firstly, we record 31 speech imagery contents, including 26 alphabets and five commonly used punctuation marks, from seven subjects using a 32-channel electroencephalogram (EEG) device. Secondly, we introduce the wavelet scattering transform (WST), which shares a structural resemblance to Convolutional Neural Networks (CNNs), for feature extraction. The WST is a knowledge-driven technique that preserves high-frequency information and maintains the deformation stability of EEG signals. To reduce the dimensionality of wavelet scattering coefficient features, we employ Kernel Principal Component Analysis (KPCA). Finally, the reduced features are fed into an Extreme Gradient Boosting (XGBoost) classifier within a multi-classification framework. The XGBoost classifier is optimized through hyperparameter tuning using grid search and 10-fold cross-validation, resulting in an average accuracy of 78.73% for the multi-character classification task. We utilize t-Distributed Stochastic Neighbor Embedding (t-SNE) technology to visualize the low-dimensional representation of multi-character speech imagery. This visualization effectively enables us to observe the clustering of similar characters. The experimental results demonstrate the effectiveness of our proposed multi-character classification scheme. Furthermore, our classification categories and accuracy exceed those reported in existing research.